Power-supply designs are often overlooked as part of a high-frequency electronic system. A wide range of approaches are used—from the simplest linear regulators to switching converters and complete distributed power systems—and many of these power designs simple don’t work as well as they should. They often fail to meet key performance requirements, such as stability, regulation, ripple, and headroom. There are a variety of technical reasons why these power supply designs perform poorly, and it may help to examine these issues individually.

In many designs, the same power supply (consisting of a converter and many regulators) is used to feed various loads, which are often digital or RF in nature (Fig. 1). But the increasing speeds of devices in these circuits can mean that the transient edges of load currents are becoming more of a problem, with fast rise and fall times. Regulators have also increased in speed, often with bandwidths in the 5 to 10 MHz region, so that a standard 2-MHz network analyzer has the capability to perform measurements on these regulators. As a result, power electronics engineers may find themselves struggling with RF engineering issues.

Often, it is an RF or system engineer assembling a distributed power subsystem for a military electronics system, typically with an off-the-shelf or generically designed DC-DC converter. The general belief is that a linear or low-drop-out (LDO) regulator is simply a drop-in part that receives an input and generates an output. There is a lack of understanding regarding the relationships between the power devices and the system—for example, how decoupling capacitors and filter capacitors affect the stability or transient performance of the voltage regulator. In space applications, large capacitors are needed to support single-event and other radiation effects. However, these capacitors drastically impact the stability of the regulator.

Power system design typically begins by identifying what voltages are needed and selecting switching regulators with multiple outputs. Rarely is a tradeoff study performed to assist with the selection of a converter topology. This is certainly not the optimum approach. Often, a flyback topology with a multi-output transformer winding is selected for the converter. This can result in very large ripple and unpredictable cross-regulation and sta-bility issues due to the intricate nature of the transformer.1-3 Analyses of these types of power supplies are difficult and often intractable. But simplified analyses can fail to uncover possible issues at certain operating points that don’t reveal themselves until they are encountered during the flight mission. At that point, the first design errors are patched with linear regulators and adequate filtering to address single-event effects (SEEs), electromagnetic interference (EMI), and other noise and ripple concerns.

One of the reasons why the multiple-output-winding flyback topology is used, as opposed to several single or dual converters that would provide better ripple performance, is its low parts count. Unfortunately, this benefit is wiped out by the large amount (in both quantity and board area) of filtering required by the post regulators.

Regrettably, voltage regulator product data sheets do not provide enough information to be useful in performing a tradeoff analysis. Regulator data sheets typically lack information regarding stability, power supply rejection ratio (PSRR) versus differential voltage, reverse transfer effects, and the impacts of adding filters. They also typically lack information on how to set the adjustment pin capacitor value or the load capacitor’s equivalent-series-resistance (ESR) range for stable regulator operation. Most regulator data sheets focus on regulation accuracy and dropout voltages. As an example, the data sheet for a model LM317 positive linear voltage regulator from ON Semiconductor (www.onsemi.com) includes this advice: “Although the LM317 is stable with no output capacitance, like any feedback circuit, certain values of external capacitance can cause excessive ringing. An output capacitance (CO) in the form of a 1.0-µF tantalum or 25-µF aluminum electrolytic capacitor on the output swamps this effect and ensures stability.” However, the typical phase margin of the LM317 application circuit with no output capacitor and a small (2-mA) load step is poor and looks nothing like the “Load Transient Response” curve in the datasheet (Fig. 2).

Why do such differences exist? For one thing, the data sheet shows a large-signal response rather than a small-signal response, so the data sheet is not a valid representation of stability. Also, the data sheet uses very slow rising and falling edges for the step load, masking any high-frequency instabilities.

A wide range of different capacitors are currently available from numerous suppliers, so a 1-µF tantalum capacitor can have an ESR that is very low or very high. Additionally, a capacitor’s ESR and equivalent series inductance (ESL) are generally uncontrolled parameters, so only maximum values are typically specified on a data sheet. The capacitor’s ESR is largely responsible for the stability of the regulator.

The measurement of regulator ripple rejection can also reveal some interesting effects (Fig. 4). When ripple rejection characteristics are given at 120 Hz, they can be misleading, along with the fact that load current optimizes ripple rejection. The data sheet mentioned above provides ripple rejection only to about 1 MHz, which is lacking a great deal of information for practical, modern applications.

The same is true of the output impedance, which is also only shown to 1 MHz (Fig. 5). The data is provided at a single operating current and limited frequency. The output impedance varies depending on load current, and the important data is actually above 1 MHz, since the bandwidth of the circuit is in this region (which corresponds to the load step response in Fig. 3). Since phase margin is largely determined by how the voltage regulator is loaded, information not included on a regulator data sheet can hinder the design effort when it is required to understand how regulator performance will be impacted by poor phase margin.

Suppliers of voltage regulator integrated circuits (ICs) do not typically provide software models for use with common RF/microwave simulators; more often, their models may support SPICE simulations. But RF/microwave design engineers don’t typically use SPICE as much as they use RF/microwave simulators, such as the Advanced Design System (ADS) from Agilent Technologies (www.agilent.com) or Microwave Office® from AWR (www.awrcorp.com), which are much more capable than SPICE for RF analysis and include many capabilities that SPICE does not, including large-signal AC simulation. An RF/microwave or system designer needs to see what the simulation results are in their own systems’ output. Therefore, regulators must be included with the RF circuitry in the system simulation. SPICE cannot run RF simulations, so in order to complete the higher-level analysis, the RF engineer needs the appropriate power IC models translated to their tools.

Another issue in developing proper power circuitry for high-frequency circuits and systems has been the lack of test equipment interface adapters and signal injectors capable of making the necessary, high-fidelity measurements. While network analyzers, impedance testers, and other test equipment with adequate resolution are available, the interface adapters used to connect a device under test (DUT) can distort the system performance to the point that measured data is erroneous.4-6 Several manufacturers now offer improved Bode injection transformers.7 But not all injection transformers are capable of the small-signal, wide-bandwidth measurements required for a good Bode plot. Some engineers believe that audio and video transformers can be used and that the transformer does not affect the measurement, but both of these beliefs are false. The bandwidth required for a proper Bode plot or impedance measurement is generally at least several octaves above the control-loop bandwidth. A typical linear regulator has a bandwidth of several megahertz, requiring high-fidelity equipment to reveal the regulator’s true performance.

A good laboratory test setup should have several types of injection transformers, each optimized for a different bandwidth. Power factor correction applications require an injection transformer with good low-frequency performance (in the 1-Hz range), while offline power supplies and LDOs require much higher bandwidths. Some state-of-the-art voltage regulators with bandwidths in the megahertz region can require a solid-state injector to properly resolve the higher-frequency information.

It should be noted that injection transformers use a very high permeability, specially annealed core material. The typical injection transformer, high quality or not, cannot operate with more than 5 to 10 mA DC current. Higher currents will provide incorrect results, but also can permanently bias the core, rendering the transformer useless and possibly damaging the driving amplifier as a result of transformer saturation. Additionally, consideration should be given to how measurements can be made noninvasively under final production conditions, as test boards, prototypes, and engineering models with a different layout won’t present the same loading and transmission line impedance as the final hardware. Noninvasive stability measurements are possible with the right equipment.

The proper tools are needed to successfully complete a design for a distributed power system. These include the right architecture and converter topology, in addition to more comprehensive datasheets, high-fidelity sim-ulation models, and support for the system-level simulators (not SPICE simulators). Accurate test equipment with sufficient measurement capability is also needed.

The road to proper power design requires a multifaceted approach. First, RF/microwave system designers must know something about power conversion so they can make intelligent decisions. RF system designers must be taught that the power system is not a black box to be inserted into the system at the end of the design process. Power requires planning.

All too frequently, power systems are plagued with stability, transient response, and ripple rejection problems due to poor phase and gain margin. Most high-reliability systems require a minimum end-of-life worst-case phase margin of 30 to 45 deg. Many regulators available today do not meet that, even initially at room temperature using the application circuit in the datasheet! A simple comparison of a voltage regulator’s data sheet with a data sheet for a video operational amplifier will show the lack of detail provided for the voltage regulator.

All too often, efforts made at saving money can backfire, such as cutting back on travel and education budgets for engineers. Lack of knowledge at the engineering level results in design errors that can be expensive and time-consuming to correct. Designers must learn how to properly use power ICs. They must be equipped with the proper tools, such as comprehensive information on product data sheets and high-quality computer simulation models. And RF/microwave engineers need to be ready to make high-fidelity measurements to support their design and analysis efforts. With improved tools, including accurate and complete data sheets, come improved power circuits. DE

References

1. Dragan Maksimoviæ and Robert Erickson, “Modeling of Cross-Regulation in Multiple-Output Flyback Converters,” Colorado Power Electronics Center, http://ece-www.colorado.edu/~pwrelect.

2. Dragan Maksimoviæ, “Computer-Aided Small-Signal Analysis Based on Impulse Response of DC/DC Switching Power Converters,” IEEE Transactions On Power Electronics, Vol. 15, No. 6, November 2000.

3. Robert W. Erickson and Dragan Maksimoviæ, “A Multiple-Winding Magnetics Model Having Directly Measurable Parameters,” Colorado Power Electronics Center, PESC98, http://ece-www.colorado.edu/~pwrelect.

4. Steve Sandler and Paul Ho, “Deconstructing the Step Load Response Reveals a Wealth of Information,” Internal White Paper, Picotest, www.picotest.com.

5. Steve Sandler and Tom Boehler, “Why Network Analyzer Signal Levels Affect Measurement Results,” Internal White Paper, AEiSystems, www.AEiSystems.com.

6. Steve Sandler and Charles Hymowitz, “Essential Test Adapters for Your Network/ Impedance Analyzer,” Internal White Paper, Picotest, www.picotest.com.

7. Steve Sandler and Charles Hymowitz “Broad Injector Product Line Eases Power Supply Stability Analysis,” Picotest, http://www.how2power.com/newsletters/1007/products/H2PToday1007_products_Picotest.pdf.

The top 10 things RF designers should know about voltage regulators

1. Low dropout is not always a good thing.
2. Most regulators can become unstable when a filter is connected.
3. Bigger is not always better: A large bandpass element has a large capacitance.
4. Not all regulator topologies perform similarly.
5. If a manufacturer does not state or claim stability performance, the device is likely not stable.
6. Always be aware of how data sheet performance values are measured.
7. Know your capacitors. ESR is not a constant, but a function of many variables. Capacitance can be very voltage sensitive, especially in the case of X5R capacitors.
8. A regulator can be effectively simulated in SPICE, ADS, Microwave Office®, or Altium Designer with good models of the regulator and output capacitor, but this will require modeling the parts yourself, hiring a modeling specialist, or convincing the IC vendor to make a high-fidelity model in the required simulator’s syntax (such as for ADS).
9. A very high-quality signal injector is needed to measure a typical regulator control loop—the measurement bandwidth usually must be much higher than expected.
10. Not all regulators allow stability measurements to be made on them. It is still possible to get an indication from the step-load performance if it is known how and where to inject the signal using a high-fidelity current-signal injector.